Date:         Wed, 20 Dec 2006 11:16:16 -0600
Reply-To:     Robin High <robinh@UNLSERVE.UNL.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Robin High <robinh@UNLSERVE.UNL.EDU>
Subject:      Re: proc mixed for 2 groups (Placebo and all Active Treatment
              Combined)
Comments: To: Jake Bee <johbee@GMAIL.COM>
In-Reply-To:  <ed52c9100612200731o3ed22ec6w78f66ee0b6686ed@mail.gmail.com>
Content-Type: TEXT/PLAIN; charset=US-ASCII

> Previous I asked this group for a proc mixed solution for Placebo vs.
> Multiple Treatment Groups and a Dunnett Test,
> and the answer was correct.  I'm not a statistician (programmer).  The
> solution was:
>
> proc mixed data=rx_1;
>  by week;
>   class center_eff treatment;
>    model rx_change_rx = rx_baseline center_eff treatment / solution;
>    lsmeans treatment / diff=control cl adjust=dunnett;
>   run;
> quit;
>

> Can this be setup for a T-Test for the 2 Treatment Groups Placebo vs.
> Active Groups (i.e. combined)

Jake,

 proc mixed data=rx_1;
 by week;
 class center_eff treatment;
 model rx_change_rx = rx_baseline center_eff treatment / solution cl;
 lsmeans treatment / diff cl;
 run;

If treatment has two groups (active and placebo) and you want to compare
two means like a t-test, then the LSMEANS statement as above will do it.

Also, the "solution" pvalue for treatment and the type 3 table from the
MODEL statement will also give you the same results. And while we're at
it:

ESTIMATE 'active vs placebo' treatment 1 -1 / cl;

will also compute it.

Other items of interest is the name of the term center_eff -- is that
factor actually considered fixed, or could it be entered as a random
effect on a RANDOM statement?

RANDOM center_eff ;

(and thus removed from the MODEL statement).

And why is

BY week;

entered?  My next question, is there repeated measures on the same
subjects over time (unless, of course, you have different subjects each
week, then ignore this Q.).

Robin High